Heparin-induced thrombocytopenia (HIT) is an acquired prothrombotic disorder caused by the anticoagulant heparin. In vivo, heparin binds to platelet factor 4 (PF4), a heparin-neutralizing protein released in high concentrations from the alpha granules of activated platelets. Large multimolecular complexes form when the two are present at optimal stoichiometric concentrations [1–4]. In some patients, the heparin-PF4 complex triggers an immune response that results in the production of an antibody, usually IgG, whose F(ab) domain binds to the heparin-PF4 complex [1, 5–7]. As these immune complexes assemble on the platelet surface, the Fcγ domains crosslink the FcγIIa receptors of nearby platelets, resulting in platelet activation [8, 9]. This leads to further release of PF4, creating a positive feedback loop that propagates platelet activation and ultimately leads to removal of these platelets from the circulation . The resultant thrombocytopenia, however, is not associated with an increased bleeding risk. Instead, ongoing platelet activation by the immune complexes results in increased thrombin production and a systemic hypercoagulable state that can culminate in venous and/or arterial thrombosis .
The immunobiology of HIT is perplexing in that the immune response is not typical of either a primary or secondary (anamnestic) immune response. In a primary immune response, IgM antibodies appear at approximately day 5 of antigen exposure, followed several days later by a relatively weak IgG response. In a secondary immune response, high titers of IgG antibodies appear very rapidly after exposure to antigen and persist in measurable amounts for several months to years thereafter, as is seen after vaccination and alloimmunization. In patients who mount an immune response to heparin-PF4, the majority form IgG antibodies, with or without concomitant IgA and IgM, between days 4 and 10 of heparin therapy [12, 13]. The simultaneous appearance of multiple isotypes within a few days of antigen exposure is consistent with an anamnestic response, but the rapid decline in IgG titers in patients with clinical HIT (becoming undetectable within 50–80 days, or sooner, after heparin is discontinued) is not in accordance with this type of immune response. Even more strikingly inconsistent is that the decline of antibody reactivity has been observed in some patients with clinical HIT despite ongoing heparin therapy . In short, the lack of IgM precedence, with evidence of previous class switch, points toward a secondary immune response, but the relatively weak and transient nature of this response is very atypical for an anamnestic immune response.
Prompt diagnosis of HIT is essential to prevent or mitigate thrombotic complications. Diagnosis of HIT can be problematic and requires consideration of both clinical likelihood (i.e., pretest probability) and laboratory data. Oftentimes, patients with suspected HIT are hospitalized and have a number of other reasons to be thrombocytopenic, including drugs, sepsis, disseminated intravascular coagulation, and prosthetic valves. A commonly used scoring system to ascertain pretest probability of HIT uses the “4Ts”: the degree of thrombocytopenia, the timing of thrombocytopenia with respect to heparin exposure, the occurrence of thrombotic complications, and the presence of alternative explanations for thrombocytopenia . The pretest probability is then combined with results of laboratory testing for HIT antibodies to predict the likelihood of HIT [16, 17].
There are two approaches to detect HIT antibodies. The most commonly used methods are immunoassays, which detect the presence of antibodies that bind to heparin-PF4 complexes without regard to the functional abilities of these antibodies. Less commonly used are platelet activation assays, such as the serotonin-release assay (SRA) and heparin-induced platelet aggregation assay (HIPA), which demonstrate the presence of a clinically relevant antibody.
Although the functional assays are generally considered the gold standard for diagnosing HIT, they are largely unavailable. These assays are time-intensive, which limits their usefulness in upfront clinical decision-making. They also require highly skilled laboratory technicians, as well as a pool of normal platelets from healthy donors, both of which are potentially costly and hard to find in smaller community medical centers. Moreover, the SRA requires the use of radioisotopes, which most clinical laboratories are now trying to avoid due to regulation and safety issues. According to a survey of 44 of the 57 laboratories that comprise the North American Specialized Coagulation Laboratory Association (NASCOLA), only 13 laboratories perform a platelet activation assay. Of these, there is an even split between laboratories that perform HIPA using platelet-rich plasma (6 of 13) and laboratories that perform SRA using washed platelets (6 of 13). None of the laboratories used washed platelets for the HIPA and only one laboratory used platelet-rich plasma for the SRA. In the HIPA assay, platelet-rich plasma from a healthy donor is incubated with the patient's serum and either collagen (positive control), buffer (negative control), therapeutic heparin (0.1–1.0 U/mL), or supratherapeutic heparin (100 U/mL). As with all in vitro tests of platelet aggregometry, a source of shear force is required, typically a magnetic stirrer. The final transparency of the reaction mixtures is directly proportional to the amount of platelet aggregation. The test is considered positive if there is platelet aggregation (high transparency) in both the positive control and therapeutic heparin concentrations, but not in the negative control or supratherapeutic heparin concentrations.
Laboratory detection of HIT antibodies
In the SRA, donor platelets are washed and incubated with serotonin that has been radiolabeled with carbon-14. During this incubation period, the platelets take up the radiolabeled serotonin and store it in their dense granules. The platelets are then incubated as described above. The supernatants of each reaction mixture are then collected and their radioactivity is measured. The test is considered positive if there is >20% release at therapeutic heparin levels and <20% release at supratherapeutic heparin levels. The SRA has been shown to be the most sensitive and specific of the functional assays [18–21], but it is time-intensive, usually a send out (obviating its practical use in the acute clinical setting), requires highly skilled personnel and specialized laboratory equipment, and a laboratory certified in the use of radioactive materials. Similarly, the HIPA assay is not readily available, but it is somewhat more practical in that it does not require radioactive materials and the turn around time is shorter, as there is no incubation with radiolabeled materials and most laboratories use platelet-rich plasma rather than washed platelets, removing an additional step. Studies suggest that improved sensitivities can be achieved by the HIPA by using a prescreened donor pool whose platelets are known to be reactive to HIT antibodies (sensitivity up to 81% with the most reactive platelets) . Likewise, the specificity of the HIPA assay can reach up to 100% by performing a “confirmatory” step with supratherapeutic heparin concentrations, which disrupt the heparin-PF4 complexes.
Although generally considered the definitive diagnostic tests for HIT, the functional assays remain impractical and inaccessible, and are therefore of very limited use in the acute setting. The enzyme-linked immunoassay (EIA) method of detecting heparin antibodies was readily adopted by specialized coagulation laboratories because of its technical simplicity and the ease with which it can be automated using equipment already available in most laboratories. Unlike the platelet aggregation assays, EIA does not require a normal source of platelets, furthering its appeal. According to the NASCOLA survey, 88% of laboratories use EIAs to detect heparin antibodies. The majority of these laboratories use commercially available kits made by either Genetics Testing Institute (GTI, Waukesha, WI) or Stago (Diagnostica Stago, Asnieres Sur Seine, France) . These assays use heparin-PF4 (Stago) or polyvinylsulfate-PF4 (GTI) immobilized on microtiter plates as target antigens. If an antibody in the patient's plasma binds to the target antigen, it is then detected by a goat-antihuman IgG/A/M that is conjugated to alkaline phosphatase. A substrate is then added to the mixture and, in the presence of alkaline phosphatase, changes color. The intensity of the color change is measured and reported as an optical density (OD).
The ELISA is ideal for ruling out HIT. Its sensitivity is very high, upwards of 95%, so persistently negative tests argue strongly against the diagnosis of HIT. The specificity, however, is quite a bit lower (74–86% or higher, depending in part on the assay manufacturer), accounting for the high rate of false positives [18, 24, 25]. Interestingly, the rate of false positive results (defined as positive serology in the absence of thrombocytopenia and other clinical criteria for the diagnosis of HIT) varies depending on the patient population. For example, in postcardiac surgery patients, the false positive rate can be as high as 50–60% [26–30]. As a result, positive tests are more challenging to interpret than negative tests. To aid in the interpretation of a positive EIA, the clinician should consider the clinical scenario (i.e., the pretest probability based on the “4Ts”) and the absolute value of the OD. Several studies have documented that high OD values (corresponding to high titers of HIT antibodies) are associated with a higher likelihood of a positive HIPA, a higher “4Ts” score (clinical HIT), and a higher probability of thrombotic complications [31–35].
There are also laboratory techniques that can be employed to further elucidate whether a detected antibody is clinically relevant. As with the functional assays, the laboratory can perform a “confirmatory” step using high concentrations of heparin that disrupt the heparin-PF4 complexes. A persistently positive test may suggest the presence of an antibody that binds to PF4, but not to the heparin-PF4 complex . In a retrospective study at Duke University Medical Center, the authors showed that patients with a positive confirmatory step were more likely to meet clinical criteria for HIT compared to patients, who had a negative confirmatory step (72% versus 18%) . Such an antibody is less likely to be of clinical relevance. That being said, in a few instances the confirmatory step misclassified true HIT plasmas with a negative confirmatory test. A more recent study showed that the heparin confirmatory step differs considerably between two commercial IgG-specific EIAs, although the recommended final concentration for “high” dose heparin also differs considerably between the two assays (2 IU/mL versus 100 IU/mL), potentially accounting for the lower rate of inhibition of their respective confirmatory steps (28/39 inhibition versus 32/33 inhibition) . The utility of the confirmatory step in the diagnosis of HIT remains controversial and needs to be prospectively confirmed. Additional controls that can be used to further characterize detected HIT antibodies include FcR-blocking monoclonal antibodies. When present, these antibodies occupy the FcR binding site on platelets, thereby inhibiting platelet activation by HIT antibodies and demonstrating their dependence on the platelet FcR.
Another method that may be of use in determining the clinical relevance of a detected antibody is rerunning the EIA using an IgG-specific antibody. Studies suggest that the Fc portion of IgG is a critical component in the pathogenesis of HIT, binding to FcγIIa receptors on platelet membranes and causing platelet activation [1, 5–9]. It is therefore less likely that an IgA or IgM antibody is clinically relevant. In fact, the frequent detection of IgM and IgA antibodies by EIA highlights two of the major drawbacks of this test. First off, EIAs cannot distinguish between platelet-activating antibodies and nonplatelet-activating antibodies. Greinacher et al. recently showed that only about 10% (17 of 185) of patients with EIA-positive sera had detectable IgM and/or IgA in the absence of detectable IgG. Of these, only one sample was positive by HIPA. The sera from these patients also failed to induce generation of platelet-derived microparticles, and only slightly increased Annexin-V binding, both markers of platelet activation. Furthermore, none of these patients developed thromboses . The second major drawback to the EIA is that it cannot detect antibodies directed against other nonheparin-dependent antigens. A small subset of HIT cases are due to antibodies against IL-8 or NAP-2 instead of PF4, and the commercially available EIA assays cannot detect these types of HIT antibodies [39, 40]. In the same study cited above, the authors showed that after IgG-depletion of two samples that were HIPA positive but EIA-IgG negative, the patients' sera became HIPA negative, suggesting the presence of platelet activating antibodies that could not be captured by the heparin-bound antigen of the EIA.
Several other studies call in to question the role of non-IgG antibodies in the pathogenesis of HIT. In a study published in 2002, the authors specifically looked at isotype breakdown of HIT antibodies in patients with suspected HIT compared with normal healthy controls. Their data was surprising in that 66% of normal human subjects (subjects not suspected of HIT with no history of heparin exposure, thrombocytopenia or thrombosis, and negative SRAs) had detectable IgM by GTI's EIA . This finding remains to be confirmed, as few studies look at the incidence of HIT antibodies in normal healthy subjects, and even fewer look specifically at isotype breakdown. For example, another study using a different EIA method (that detects IgG, IgM, IgA) found only 1% of healthy controls tested positive . Suh et al. showed that patients with a positive SRA and clinical HIT were significantly more likely to have detectable IgG antibodies than patients with thrombocytopenia who were SRA negative (100% versus 46%, respectively). Conversely, the non-HIT group had a much higher incidence of detectable IgM antibodies compared with the HIT group (81% versus 42%) . These findings suggest that circulating non-IgG antibodies to heparin/PF4 are commonly found in normal subjects and are similarly prevalent in patients suspected of HIT, making it difficult to assign clinical relevance to such antibodies.
The key role of IgG in the pathogenesis of HIT is becoming increasingly evident, prompting investigations into the performance characteristics of IgG-specific EIAs compared with poly-specific EIAs (Poly-EIAs), which detect antibodies of all isotypes (IgM, IgA, and IgG). Presumably, if IgG is the principal mediator of HIT, then an IgG-specific EIA could retain high sensitivity while achieving a higher specificity since it will not detect the less clinically relevant IgA and IgM antibodies that bind the heparin-PF4 antigen. In a recent prospective study comparing the positive and negative predictive values of Poly-EIA with IgG-specific EIA in a cohort of 500 patients, Bakchoul et al. showed a positive predictive value (PPV) of 40% for IgG-specific EIA versus 28% PPV for the Poly-EIA. These values were calculated based on clinical scoring (the 4Ts) and HIPA results. The sensitivities and negative predictive values for both tests were 100% using a cutoff OD of 0.4. The specificity was higher for the IgG-specific EIA (89% versus 81%), and could be further increased by increasing the OD cutoff at the expense of sensitivity, as would be anticipated. The authors found that a cutoff OD of 0.65 reflected the optimal compromise between sensitivity and specificity (sensitivity 97.1% and specificity 92.4%) . This cutoff cannot be universally applied, however, as OD thresholds vary between laboratories. More recently, Warkentin et al. showed that EIA specificity may be improved up to 89% or 99% (depending on the assay manufacturer) using IgG specific assays, with a possible corresponding slight decrease in sensitivity . Ultimately, the most definitive way to determine the clinical relevance of an H-PF4 antibody detected by ELISA is to perform one of the functional tests (HIPA or SRA), but the clinician must keep in mind that false negative results can occur with any method, and clinical context is a key part of the diagnosis of HIT.
There has been a concerted effort to develop technically simpler assays with rapid turnaround time for the detection of HIT antibodies. To this end, a group in Europe developed a particle gel immunoassay (PaGIA) based on the ID-Micro Typing System (DiaMed AG, Cressier sur Morat, Switzerland) widely used in blood bank serology. This system utilizes high-density polystyrene beads that are dyed red and coated with heparin-PF4 complexes. Heparin-PF4 antibodies from the patient's serum or plasma bind to the beads. An antihuman antibody is then added to agglutinate the beads that have heparin-PF4 antibodies adhered to them. Upon centrifugation, the agglutinated beads do not pass through the gel, forming a red band at the top of the gel; while unbound high-density beads pass freely through the gel, forming a red band at the bottom of the test tube. A positive result (red band at the top of the test tube) is easily visible without the need for specialized laboratory machinery.
To date, few prospective studies have compared the performance of PaGIA to ELISA in predicting HIT. In the aforementioned study by Bakchoul et al , the authors included PaGIA in their evaluation of different antigen binding assays. The PaGIA had a PPV of 36.6%, compared with 28% for Poly-EIA and 40.6% for IgG-specific EIA. The NPV, however, was only 99.5%, compared with 100% for the other two tests. PaGIA failed to detect two patients with a positive HIPA and intermediate to high pretest probabilities. A similar problem with false negatives was seen in a study by Pouplard et al. , in which the PaGIA failed to detect HIT antibodies in a patient with a positive SRA and EIA. The NPV in their study for PaGIA was 99.4%, compared with 100% for the SRA and 100% for the 4Ts clinical score. When used in combination with the 4Ts score, however, the authors showed that PaGIA could reliably rule out HIT. In fact, in patients designated as “intermediate risk” by clinical scoring, a negative result from the PaGIA reduced the probability of HIT to 0.6%. Schneiter et al. verified reproducibility of PaGIA results between plasma and serum specimens, fresh and frozen samples, and different polymer lots. Their results show significant agreement between plasma/serum and fresh/frozen samples, but a striking variability of false negative results depending on the polymer lot used. They suggest titrating commercial positive controls and stored samples from HIT patients to identify such faulty lots .
According to the ECAT 2009 HIT survey, which is a proficiency program that assesses a laboratory's ability to correctly test unknown specimens, 27 of 120 participating laboratories use the PaGIA. The assay performed well, with 26 of 27 laboratories getting a negative result with the negative control provided by ECAT (the remaining laboratory getting a “borderline” result), and 25 of 27 getting a positive result with the positive control provided by ECAT (the remaining two laboratories getting a “borderline” results). These 2009 results are improved over the initial proficiency survey in 2007, in which 3 of 32 laboratories gave false negative PaGIA results. The advantages of PaGIA are technical simplicity and rapid turn around time, and while most clinical experience reported to date utilizes EIA methods to detect heparin-PF4 antibodies, PaGIA may represent a simple point of care alternative. More data comparing the performance of these methods over a range of antibody titers and correlating with clinical outcome would be valuable.
PaGIA beads can also be used for flow cytometric detection of HIT antibodies. Tazzari et al. demonstrated this by incubating heparin/PF4-coated high-density beads in patient serum, and subsequently in fluorescein isothiocyanate (FITC) conjugate rabbit antihuman IgG . The beads were then run through a flow cytometer, gated by physical parameters, and their fluorescence analyzed. The results of the flow cytometric screen for HIT antibodies correlated strictly with EIA and PaGIA: 30/30 HIT patients tested positive by EIA, PaGIA, and flow cytometry; 0/10 non-HIT patients with thrombocytopenia and 0/30 normal healthy subjects tested negative by all three assays. Gobbi et al. took the application of flow cytometry a step further by combining the antibody screen described above with a functional assay using donor platelets . In their study, donor platelets were incubated with HIT patient serum in a second test tube. The donor platelets were then labeled with anti-CD41 phycoerythrin (PE) monoclonal antibodies, which recognize the glycoprotein IIb/IIIa complex on the surface of platelets. The mixture was subsequently incubated with patient serum followed by Annexin V FITC, which binds exposed phosphatidyl serine on activated platelets. The contents of the two test tubes (the microbeads + FITC anti-IgG mixture and the donor platelet/anti-CD41 + Annexin V FITC mixture) were then combined in a single test tube and evaluated by the flow cytometer. By plotting side scatter versus FITC fluorescence, platelets and beads were easily distinguished. The degree of fluorescence of both Annexin V by the donor platelets and anti-human IgG FITC by the Heparin:PF4 beads were analyzed simultaneously. They compared their results against the clinical setting and standard EIA results of 13 patients with suspected HIT and six normal healthy subjects. In seven of the thirteen patients with suspected HIT, the results of flow cytometry were in concordance with the EIAs (four positives and three negatives). Two of the patients with discordant results were lupus anticoagulant positive with only weakly positive EIAs. The remaining four cases had frankly positive EIAs but the diagnosis of HIT was not confirmed. Of note, the results of all samples run through the flow cytometer were internally consistent, either positive for both platelet activation and HIT antibodies or negative for both. Not included in their study, but certainly of interest, is how the functional component of their assay compares with the traditional functional assays like SRA and HIPA.
Flow cytometry appears to be a promising alternative for the rapid screening of HIT. It is also the first laboratory test in which both the presence of an antibody and its functional abilities can be detected simultaneously. The procedures described above are technically straightforward and fast, making flow cytometry an attractive possibility. Larger prospective studies are still needed to confirm the clinical utility of flow cytometry in the diagnosis of patients with suspected HIT.
There are a handful of other immuno- and functional assays, including fluid-phase EIAs, ATP-release chemiluminescence assays, flow cytometric detection of platelet microparticles, and thrombin generation studies that have been investigated in the research setting only, and like the above described flow cytometric studies, have yet to be validated prospectively in a clinical setting [11, 47–50]. The commercially available EIAs remain the most widely used assays in the clinical diagnosis of HIT. When the results of the EIA are equivocal and the clinical setting is questionable, some laboratories will send out or perform one of the traditional tests of platelet activation, either SRA or HIPA, to confirm or refute the diagnosis. One caveat of this approach is that occasionally, true cases of HIT can have a positive EIA with a negative SRA or HIPA. Despite their wide use, the preanalytic handling, testing methodologies, and results interpretation and reporting of the most commonly used assays, including the commercially available immunoassays and the “home brewed” platelet-activation assays, is highly variable among the specialized coagulation laboratories, and presumably even more so among unspecialized laboratories  (Figs. 1 and 2). There is a clear need for standardization of procedures for the functional assays, as well as proficiency testing, as the actual performance characteristics of the functional assays are not known outside of the research setting. There is a similar need for positive control material to validate the functional assays, irrespective of the instruments or reagents used, and to aid in the identification of optimal donors whose platelets are sensitive to HIT antibodies.
HIT remains a clinicopathologic disorder requiring the presence of one or more clinical features, usually thrombocytopenia and/or thrombosis in the setting of timely heparin exposure, as well as laboratory detection of pathogenic HIT antibodies. The EIA is a highly sensitive assay, but its specificity is considerably lower and varies depending on the patient population and type of heparin used [51–53]. Of all patients exposed to heparin, only a subset will have a positive EIA, and only a fraction of that subset will make antibodies capable of activating platelets in the SRA or HIPA. Fewer still will develop evidence of clinical HIT (thrombocytopenia and/or thrombosis). This phenomenon is well illustrated by the Iceberg Model of HIT developed by Warkentin . This model applies to all types of sulfated polysaccharide anticoagulants, including unfractionated heparin and low molecular weight heparin. Unfractionated heparin is the more immunogenic and antigenic, and is therefore depicted by a larger iceberg (corresponding to immunogenicity) with a larger proportion protruding above the waterline (corresponding to antigenicity). To better predict which patients are at highest risk for clinical HIT, the pathophysiology underlying the Iceberg structure must be delineated. As these mechanisms unfold, the diagnosis of HIT will become more definitive as the screening and confirmatory tests become more refined. Until then, physicians must rely on their clinical judgment on a case-by-case basis, incorporating their knowledge of the clinical scenario with the strengths and weaknesses of the laboratory tests available to them.
Clinicians should be aware that testing for HIT cannot be performed after the acute setting because the antibody gradually disappears after heparin exposure is discontinued, and the HIT test will likely be negative even if the patient had true HIT. We propose an algorithm to aid clinicians in the acute setting with the diagnosis and management of patients with suspected HIT (Fig. 3). These guidelines are meant to assist the clinician, but should not trump clinical judgment in individual cases. Several other testing algorithms have been proposed and are also suitable.